Invited Perspective: Cutting through the Noise—the National Park Service Anthropogenic Noise Model for Exposure Assessment Authors: Danielle Vienneau, Jean Marc Wunderli Published: 2023 Noise—unwanted sound—is a major yet often ignored environmental risk factor affecting the health and well-being of large numbers of people, especially in cities, where transportation noise is a major source. With urbanization and an aging population more susceptible to chronic diseases, we can expect noise exposures and the associated health burden to increase. As a psychosocial stressor, and often beyond the control of the individual, noise can induce an emotional toll and trigger a physiological stress response,1,2 leading to effects in adults ranging from annoyance and sleep disturbance to manifestations of metabolic and cardiovascular disease (CVD).3 The 2018 Environmental Noise Guidelines for the European Region rated as “high” the quality of evidence that road traffic noise increases the risk of ischemic heart disease.4,5 At the time, data from large cohorts were sparse and limited in terms of outcomes. Much more data are now available, although still with relatively fewer studies from North America. Although the lagging research may partly relate to priorities, the major challenge for noise and health studies lies in exposure assessment when the study population spans large geographic areas, such as the whole of the United States. Therefore, it is highly commendable that a rigorous study from the United States is now available. Conducted within the Nurses’ Health Study, the study by Roscoe et al.6 in this issue of Environmental Health Perspectives is the first longitudinal study to use national modeled noise data from the U.S. National Park Service (NPS), finding increased risk of CVD and coronary heart disease incidence in association with both nighttime and daytime noise. Casting a critical eye on the exposure data is prudent, however, given that these will no doubt be used as a resource in noise and health research going forward. Originally designed to quantify sound intrusion from anthropogenic sources (e.g., road traffic, aircraft, industry, and other manmade sounds) into parks and evaluate effects on nature and wildlife, the geospatial NPS model fundamentally differs from common noise modeling approaches used in Europe: It relies on extensive measurement data plus inputs on environmental factors, such as topography, climate, hydrology, population density, and anthropogenic activity, to estimate noise levels in a machine learning approach. In contrast, European approaches aim at representing the physics of sound generation and propagation, modeling each source separately. They start with infrastructure and operational data such as traffic volumes and travelling speeds that yield a sound emission per source. The subsequent sound propagation modeling takes topography, land use categories, and building infrastructure into account and results in sound exposure levels for all buildings. Although an impressive number of measurement hours have been recorded to underpin the NPS model, independent validation is lacking, and <30% of measurement sites represent urban environments where noise levels are high and the number exposed is large.7 Such a geospatial noise model, in our view, can be valid for rural areas with small building density and for community-level exposure estimates. For specific exposure locations in cities (e.g., residential addresses), however, it is crucial to know whether the line of sight (and thus of sound) to a nearby highway is blocked by another building. In addition, detailed information about traffic infrastructure and corresponding traffic volumes are mandatory to achieve a certain level of accuracy. Danish studies, for example, now include exposure at the least-exposed building façade as a proxy for bedroom noise given the importance of sleep and the role of chronic sleep disturbance in contributing to adverse health effects of noise, such as CVD.8–11 In Switzerland, our study showed that even small aggregations in the spatial scale of the noise estimates—for example, from the façade point level to the average level in a rasterized noise map—can attenuate myocardial infarction mortality associations.12 In addition, at least for ground-level sources such as road or railway traffic, the sound exposure decreases rapidly with increasing floor height. Consequently, it would be preferable to assign the inhabitants of multistory buildings to a specific floor and conduct floor-specific exposure modeling.12 To interpret the findings from the Nurses’ Health Study within the broader noise and health context, it is necessary to understand what the noise metric (L50, the noise level that is exceeded 50% of the time) and anthropogenic (as a combined source) truly mean in the NPS model. Roscoe et al. chose to use L50 because the NPS model isolated anthropogenic noise from natural sounds for this particular metric. Most studies now use energy-based metrics to express time equivalent sound levels (Leq), such as the LNight (the A-weighted long-term average sound level determined over all the night periods of 1 y).13 For road traffic noise, L50 and the Leq are highly correlated. L50 is well suited to characterize road traffic noise owing to the typical constant flow of numerous vehicles, but it is insufficient for quantifying levels from intermittent noise sources, such as aircraft, trains, or revving motorbikes. Such intermittent and potentially loud events will not alter L50 because their duration is clearly <50% of the time, but they can have a significant influence on the Leq. Thus, neglecting infrequent, loud events seems problematic, given that these disproportionately contribute to noise annoyance14,15 and specifically sleep disturbance.9,16,17 A survey in Switzerland, for example, showed intermittency of the noise source to modify the exposure–response relationship for LNight and self-reported sleep disturbance.18 It must be assumed that the primary anthropogenic noise source captured in the NPS L50 metric is typical road traffic noise. Focus on the nighttime period seems most valid because L50 only makes sense for exposure situations with a broadly constant level. Given the strong diurnal pattern of anthropogenic noise and the longer duration of day compared to night hours, an L50 for the entire 24-h period would in fact be fully dominated by daytime noise. The nighttime focus further aligns with the many studies showing the negative effects of disturbed sleep.10,19–22 Despite a potentially high degree of exposure measurement error, especially within cities, the paper by Roscoe et al.6 provides important information from a strong nationwide cohort that anthropogenic noise is a risk factor for CVD. How exactly the findings might feed into local health-based policies or guideline development remains to be seen. Refers to https://doi.org/10.1289/EHP12906 References